Regression: kernel 5.4 does not show NVME SSD

Regression: kernel 5.4 does not show NVME SSD

[Ingo Brunberg]

Hi experts,

starting with linux kernel 5.4 my NVME SSD from HP is no longer detected
correctly, nvme list shows no result. I made no relevant changes
regarding .config between kernels 5.3 and 5.4.

Please have a look at this bugzilla entry:
https://bugzilla.kernel.org/show_bug.cgi?id=205679

Maybe all SSDs with a Silicon Motion 2263 controller are affected.
If you need more information, please ask.

Regards
Ingo Brunberg

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: Regression: kernel 5.4 does not show NVME SSD

[Keith Busch]

On Mon, Dec 02, 2019 at 10:31:18AM +0100, Ingo Brunberg wrote:
> Hi experts,
> 
> starting with linux kernel 5.4 my NVME SSD from HP is no longer detected
> correctly, nvme list shows no result. I made no relevant changes
> regarding .config between kernels 5.3 and 5.4.
> 
> Please have a look at this bugzilla entry:
> https://bugzilla.kernel.org/show_bug.cgi?id=205679
> 
> Maybe all SSDs with a Silicon Motion 2263 controller are affected.
> If you need more information, please ask.

It's a regression from commit 538af88ea7d9de241e6b6f006e9049c4d96723bb.
This one should have only returned the error if < 0, otherwise we can
proceed. Will get post a patch shortly with a 5.4 stable request.

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: Regression: kernel 5.4 does not show NVME SSD

[Sagi Grimberg]

>> Hi experts,
>>
>> starting with linux kernel 5.4 my NVME SSD from HP is no longer detected
>> correctly, nvme list shows no result. I made no relevant changes
>> regarding .config between kernels 5.3 and 5.4.
>>
>> Please have a look at this bugzilla entry:
>> https://bugzilla.kernel.org/show_bug.cgi?id=205679
>>
>> Maybe all SSDs with a Silicon Motion 2263 controller are affected.
>> If you need more information, please ask.
> 
> It's a regression from commit 538af88ea7d9de241e6b6f006e9049c4d96723bb.
> This one should have only returned the error if < 0, otherwise we can
> proceed. Will get post a patch shortly with a 5.4 stable request.

Should we have something that is more granular rather than any positive
error?

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: Regression: kernel 5.4 does not show NVME SSD

[Keith Busch]

On Fri, Dec 06, 2019 at 02:44:23PM -0800, Sagi Grimberg wrote:
> > This one should have only returned the error if < 0, otherwise we can
> > proceed. Will get post a patch shortly with a 5.4 stable request.
> 
> Should we have something that is more granular rather than any positive
> error?

I don't think so, otherwise we'll be chasing new errors from disfunctional
controllers. As long as we get a response of any kind, we can use the
controller. It's only when we don't get a response, which is what the
negative return codes are for.

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme